Transverse force sensing with a uniform FBG and unpolarized light via machine learning.

In this paper, we report that a uniform fiber Bragg grating (FBG), interrogated with broadband light, can be used to reliably measure transverse force through the analysis of its unpolarized reflected amplitude spectrum. Unlike conventional FBG-based transverse force sensors, which typically rely on birefringence-induced spectral splitting or polarization-resolved interrogation schemes, our approach leverages the full spectral shape of the reflected signal, recorded with standard illumination. Two types of measurement devices are used for the readout process: (i) a broadband light source combined with an optical spectrum analyzer, and (ii) a spectrometer-based FBG interrogator. Both approaches yield similar performance in controlled conditions, confirming the robustness of the FBG interrogator, which is the only configuration suitable for deployment in the field. By training a model with hundreds of experimental spectra measured for calibrated transverse force values between 0 and 70 N, we achieve highly accurate and robust prediction of the applied transverse force, with a mean absolute error of 2.80 N for the multilayer perceptron (MLP) and 0.44 N for the gradient boosting (XGBoost) model. This data-driven method removes the need for polarization control or complex grating structures, and opens a new pathway towards simple, cost-effective, and highly sensitive transverse force sensing.